Simple multi-dataset detection
This addresses the challenge of inconsistent taxonomies across datasets for researchers and practitioners in computer vision, offering a scalable solution for multi-dataset detection.
The paper tackles the problem of building a general object detection system by training a unified detector on multiple datasets with inconsistent taxonomies, resulting in a learned taxonomy that outperforms expert-designed ones and generalizes to unseen datasets without fine-tuning.
How do we build a general and broad object detection system? We use all labels of all concepts ever annotated. These labels span diverse datasets with potentially inconsistent taxonomies. In this paper, we present a simple method for training a unified detector on multiple large-scale datasets. We use dataset-specific training protocols and losses, but share a common detection architecture with dataset-specific outputs. We show how to automatically integrate these dataset-specific outputs into a common semantic taxonomy. In contrast to prior work, our approach does not require manual taxonomy reconciliation. Experiments show our learned taxonomy outperforms a expert-designed taxonomy in all datasets. Our multi-dataset detector performs as well as dataset-specific models on each training domain, and can generalize to new unseen dataset without fine-tuning on them. Code is available at https://github.com/xingyizhou/UniDet.